Why do so many audiophiles and respected sound engineers insist upon recording sounds above 20khz?
I think they're speculating, and they have no choice in the market as it's very hard to get something that doesn't record at a high-res rate anyway.
There have been lots of theories put forward, but no good experiment has been done to prove the efficacy of 20+ kHz recording. There have been lots of theories disproved, like we need to hear 20+ kHz signals.
One of the leading explanations (in need of someone to do a good experiment for it) is that the filters used for higher sampling rates can be made shallower, and therefore are less intrusive on the audible part. In other words, it's not that we can hear sounds above 20 kHz, but what we do above 20 kHz has consequences on what we can hear below 20 kHz.
How the filters do this has its own subset of theories. One is the very popular pre-echo explanation. Another is that the sharpened impulse response of a hi-res system gives you better imaging. Another is that the top critical band of the ear, which is non-linear, can interact with a brickwall filter's very narrow transition band so you can hear the effects of ripple caused by the brickwall filter. A higher sampling rate can relax the brickwall filter so this interaction doesn't happen.
Two interesting side notes on this: digital limiters and compressors actually benefit greatly from higher sampling rates. They are non-linear processes that produce harmonics above our hearing range, but can alias lower sampling rates like 44.1 kHz. Oversampling these processes actually makes for a nicer sound because you don't get the very unnatural aliasing distortion when the harmonics wrap around to lower frequencies.
Tom Stockham, one of the pioneers of digital recording in the 70s, designed the Soundstream system, used in so many early Telarc CD and LP releases, used a 50 kHz sampling rate because, many believe, of the filter slope issue above. However, he never published anything explaining why, but did have conversations with people about it. Stockham was an engineer's engineer, and probably had really good reasons with good experimental data to back up his decisions. It's too bad he died before anyone could ask him about it.